631
DSLs Enable Reliable Use of LLMs
Why Domain-Specific Languages and good abstractions make LLM code generation reliable and reviewable.
New
The latest blog posts to take off on Hacker News. 6,325 posts.
632
I tricked Claude into leaking your deepest, darkest secrets
How I tricked Claude into leaking your deepest, darkest secrets
633
You only need the frontier model for one single edit
/prewalk hands the cheap model a trajectory, not a plan, retaining 92–97% of frontier pass rate at roughly half the cost on SWE-Bench Pro.
634
LeMario: Training a JEPA World Model on Super Mario Bros
A from-scratch JEPA world model trained on Super Mario Bros. to study action-conditioned prediction and reward-free planning.
635
Bonsai 27B: A 27B-Class model that runs on a phone
Today we're announcing Bonsai 27B, our multimodal flagship: ternary at 5.9GB for laptops, 1-bit at 3.9GB for an iPhone 17 Pro, with a 262K-token context.
636
The Tower Keeps Rising
Vibecoding and the possible collapse of a shared language.
637
Are we offloading too much of our thinking to AI?
Reflections on autonomy and the value of thinking for ourselves
638
639
How to stop Claude from saying load-bearing
Hack the text output of Claude Code to make life a little bit sillier.
640
The Future Worth Building Is Human
AI built for autonomy crowds people out, making us passive observers of what’s coming. We’re building toward a different future.
641
What will be left for us to work on?
My keynote at ICML 2026
642
The real prices of frontier models
Real token cost of frontier models on the same TypeScript input - Claude Fable 5.1, Opus 5, GPT-5.6, Gemini 3.7 Flash. Tokenizer counts, effective cost.
643
Latent Space as a New Medium
Lately I’ve been asking myself: what might artificial intelligence be good for besides answering questions and writing code?
644
Vāgdhenu: A Sanskrit Chanting TTS System
A vṛtta (meter) aware śloka-to-chant system for Sanskrit. Try the live demo, hear sample chants, read the paper.
645
Apple's new SpeechAnalyzer API, benchmarked against Whisper and its predecessor
Compare Apple's new SpeechAnalyzer API to Whisper and SFSpeechRecognizer. Real accuracy benchmarks on 5,559 utterances with raw transcripts for rescoring.
646
I love LLMs, I hate hype
I think from this blog you may misunderestimate how absolutely giddy I am about AI. I did hacking from 2007-2014, after that my whole career has been devoted to AI. I love the progress. I’m so excited for the new LLMs, self driving cars, video generation models, and coding agents. I set up a Linux box with opencode on…
647
Claude Code sends 33k tokens before reading the prompt; OpenCode sends 7k
Claude Code vs OpenCode token overhead measured at the API boundary. Out-of-the-box baselines, instruction file weight, MCP schema tax, subagent multipliers, and cache-write behaviour.
648
Migrating a production AI agent to GPT-5.6: 2.2x faster, 27% cheaper
For four months, no frontier model beat Claude Opus in our production evals. GPT-5.6 did. This is what we learned while migrating.
649
650
Old and new apps, via modern coding agents
I have been interested in machine-assisted ways to do and teach mathematics from as far back as 1999, when I started coding several applets in Java 1.0, both for my complex analysis and linear alge…
651
Mesh LLM: distributed AI computing on iroh
How Mesh LLM pools existing GPU resources across machines into a single OpenAI-compatible API, built on iroh.
652
Stop Telling Me to Ask an LLM
I already did.
653
AI 2040 and the cult of intelligence
I used to be one of these people. I read Yudkowsky and was like, OMG recursive self improvement hard takeoff AI is coming. Then I joined the real world and actually tried to do things. At comma, we ship a hardware product of similar complexity to a cell phone, and it’s really hard. Reality has lots of finicky details…
654
Reverse centaurs are the answer to the AI paradox (2025)
My latest Locus column is "Reverse Centaurs," and it sets out to unravel a paradox: how is it that some AI's users describe their experience as a hellish ordeal, while others delight in the ways that AI is changing their lives for the better?
655
Nvidia, CoreWeave, and Nebius: Inside the Circular Financing of the GPU Boom
Neoclouds are one of the more hotly debated AI business models, with CoreWeave and Nebius being the two most widely recognized names. These companies have seen their sales, backlog, and share prices soar. Yet, supporting their growth is extremely expensive, and neoclouds do not have the same cash nor operating cash…
656
GPT-5.6, Grok 4.5, Claude, and Muse Spark build the same 4 apps
GPT-5.6's new Sol, Terra, and Luna tiers go head-to-head with Grok 4.5, Claude, Meta's Muse Spark, and the open-weights crew on a raycaster, a Rubik's cube, a calculator, and Game of Life. Here's every build, with cost and latency.
657
How the terrorist group Boko Haram uses frontier AI
The Cambridge Programme on AI Science & Policy (CASP) is an interdisciplinary research programme on frontier AI at the University of Cambridge.
658
Building a real-time AI tutor for 5-year-olds
We set out to build the first AI tutor to teach math and reading to kids ages 4-9. For AI to actually teach a five-year-old, pedagogy must be baked into the engineering. A child can't wait for a slow reply, can't read a chat interface, and can't unhear anything a model gets wrong. We wanted to share some of the…
659
GPT-5.6
More intelligence from every token, stronger performance per dollar, and more capability on demand for your hardest work.
660
ChatGPT Work
ChatGPT Work is an agent that can take action across your apps and files, stay with a project for hours if needed, and turn a goal into finished work.